Associations between a metal mixture and infant Negative Affectivity: Effect modification by prenatal cortisol and infant sex

Abstract

In-utero exposures interact in complex ways that influence neurodevelopment. Animal research demonstrates that fetal sex moderates the impact of joint exposure to metals and prenatal stress measures, including cortisol, on offspring socioemotional outcomes. Further research is needed in humans. We evaluated the joint association of prenatal exposures to a metal mixture and cortisol with infant negative affectivity, considering sex-differences. Analyses included 226 (29% White, Non-Hispanic) mother-infant pairs with data on exposures and negative affectivity assessed using the IBQ-R in 6-month-olds. Results showed that girls whose mothers had higher cortisol had significantly higher scores of Fear and Sadness with greater exposure to the mixture. Examining higher-order interactions may better elucidate effects of prenatal exposure to metals and cortisol on socioemotional functioning.

Internalizing problems, most prominently anxiety and depressive disorders, affect more than 400 million people globally (Kunas et al., 2021), with clinical and subclinical manifestations resulting in significant morbidity (Burstein et al., 2014). Rooted in childhood, there is a sharp rise in these disorders in the transition to adolescence (Kessler et al., 2007). Theories on early life origins of psychopathology and known co-morbidity across anxiety and mood disorders support a rationale for examining environmental influences through a developmental lens to elucidate their shared early patho-etiology (Melton et al., 2016). Research has largely focused on the role of early life adversity or psychosocial factors (Cameron et al., 2017), with a recent call to consider the role of environmental chemicals (Rokoff et al., 2022; Tien et al., 2020).

Temperament encompasses individual variations in behavioral tendencies of emotional responses and reactions to internal, social, and environmental stimuli and comprise various domains that reflect relatively stable traits over an individual’s lifespan (Rothbart, 2007). Infant temperament influences later personality and social development and risk for emotional and behavioral problems (Gartstein & Rothbart, 2003). While the association between infant temperament and later psychopathology is robust, the magnitude of effects can be variable (Abulizi et al., 2017; Kostyrka-Allchorne et al., 2020; Sayal et al., 2014; Slagt et al., 2016; Tang et al., 2020; Wu et al., 2022). This may in part be due to how different children respond to similar environmental challenges in predictably divergent ways, with the individual characteristics of the child influencing pathways to adaptive or maladaptive outcomes. Negative affectivity, one aspect of infant temperament, in particular has been linked to persistent difficulties including internalizing problems (Enlow et al., 2017; Gartstein et al., 2012). Moreover, research has shown that individual subscales that belong to the overarching temperament trait of negative affectivity can differentially predict emotional and behavioral problems over time (Gartstein & Hancock, 2019), with several components (e.g., sadness, falling reactivity, and discomfort) contributing to internalizing problems (Gartstein et al., 2012). Interestingly, sadness has been associated with both greater internalizing and externalizing problems (Gartstein et al., 2012). In general, heightened negative affectivity in early childhood has been shown to predict internalizing disorders in adolescence and later life (Klein et al., 2012), highlighting the need to explicate potentially modifiable environmental risk factors that contribute to early life behavioral domains. Hence, interventions can be applied early to promote optimal development and have significant implications for the prevention of chronic psychopathology.

Environmental exposures during pregnancy, particularly those known to influence brain development, may influence temperament outcomes in infants (Takegata et al., 2021). Studies to date have largely considered maternal stress and prenatal psychological functioning (Enlow et al., 2017; Van den Bergh et al., 2020) and exposure to substances, including tobacco, alcohol, and other drugs (Froggatt et al., 2020; Guille & Aujla, 2019). Prenatal exposure to a number of metals has been associated with a range of adverse neurodevelopmental outcomes, with lead (Pb), arsenic (As), and cadmium (Cd) being the most well-established neurotoxicants (Sanders et al., 2015; Shah-Kulkarni et al., 2020; Wang et al., 2018). Emerging evidence suggests that cesium (Cs), barium (Ba), and antimony (Sb), also influence behavioral development, although data are sparse (Liu et al., 2022; Nozadi et al., 2021). Much of the research on the central nervous system (CNS) effects of metals has focused on cognition/intelligence. Recent studies have examined associations with externalizing behaviors (e.g., conduct disorder), but far less work has examined how metal exposures impact emerging internalizing behaviors (Horton et al., 2018). Moreover, although metals have been linked to negative emotionality in older children, effects on behavioral functioning in early infancy has not been well studied (Fruh et al., 2019; Rocha & Trujillo, 2019).

Research has begun to link prenatal metal (de Water et al., 2018; Roy et al., 2009) and stress exposures (Klein et al., 2012; Lipton et al., 2017) with antecedents of internalizing symptoms, such as temperament. However, these have largely considered one metal or stress measure at a time. Further, elevated maternal cortisol levels in pregnancy, a biomarker of chronic stress, has also been linked with negative affectivity at 6 months of age (Enlow et al., 2017; Moisiadis & Matthews, 2014) and a recent publication showed the associations between a prenatal metal mixture and infant negative affectivity at the same age (Cowell et al., 2021). Starting in utero, toxic metals and stress disrupt similar but not completely overlapping neural processes that may underlie their interactive effects upon an unfolding predisposition to internalizing problems as children grow up – such as disrupted synaptic transmission (Dickerson et al., 2020), dopaminergic systems (Jones & Miller, 2008), and/or stress response systems, including the maternal-fetal hypothalamic-pituitary-adrenal (HPA) axis (Enlow et al., 2017).

In addition, animal studies demonstrates that infant sex may further moderate the impact of metals and stress exposures on socioemotional outcomes, although whether males or females are more impacted varies based on the outcome being assessed (Cory-Slechta et al., 2017; Singh et al., 2018; Sobolewski et al., 2018). Similarly, the current human literature also demonstrates sexually dimorphic effects in females versus males vary based on the neurodevelopmental outcome being examined. Further, no human study to our knowledge has examined associations between metals mixtures and infant temperament when considering maternal cortisol and infant sex as effect modifiers. Examining effects of these often co-occurring chemical and non-chemical exposures may also help elucidate health inequities. Pregnant women, particularly ethnic minority women living in low-income communities, often experience greater levels of traumatic and chronic stress (Evans et al., 2021) and disrupted physiological stress responsivity (Schreier et al., 2016) as well as being more highly exposed to toxic metals in their environment (Geron et al., 2022).

To extend the animal work demonstrating three-way interactions among prenatal metals, cortisol exposure, and offspring sex in relation to neurobehavioral outcomes, we leveraged a longitudinal, ethnically mixed urban pregnancy cohort to evaluate the joint association of prenatal exposure to a metal mixture (As, Ba, Cd, Cs, Cr, Pb, and Sb) and third trimester maternal cortisol assessed in maternal hair on infant negative affectivity, while considering the sexually dimorphic nature of these associations. In this confirmatory study, we hypothesized that the effect of the metal mixture on negative affectivity would be enhanced by increased cortisol, greater exposure in both being associated with greater negative affectivity in 6-month-olds. Further, these associations would vary in a sex-specific manner. Beginning in utero, a network of interconnected cells (e.g., neurons) form in the brain which stretch across different anatomic regions and several structural components of this network can be differentially susceptible to metals as well as stress and its biological correlates (cortisol) to impact a range of temperamental features (Lavenex & Banta Lavenex, 2013; Tau & Peterson, 2010). Thus, we consider both the overarching domain of negative affectivity as well as each subdomain separately in these analyses.

Methods

Study participants

The mother-infant pairs included in these analyses are participants in an ongoing prospective pregnancy cohort study – Programming of Intergenerational Stress Mechanisms (PRISM) - designed to examine the effects of prenatal and early life psychosocial, physical, and chemical environmental exposures on child developmental outcomes. Beginning in 2011, pregnant women were recruited from prenatal clinics in Boston and New York City. Women were eligible if they were English or Spanish speaking, 18 years or older, and pregnant with a singleton. Exclusion criteria included maternal intake of ⩾7 alcoholic drinks per week prior to pregnancy, any alcohol intake after pregnancy recognition, and HIV+ status. The study has recruited n=1068 women receiving prenatal care from the Beth Israel Deaconess Medical Center and East Boston Neighborhood Health Center in Boston, MA (from March 2011–December 2013) and Mount Sinai Hospital in New York City, NY (from April 2013–March 2021) who delivered a live newborn with no significant congenital anomalies noted at birth that could impact participation in follow-up procedures. Metals and creatinine were measured in a subset who also had urine samples collected during pregnancy (N=654). Among these 654 participants, infant temperament was evaluated at age 6 months for 448 infants; 229 mothers of infants who completed the 6-month assessment also provided a third trimester hair sample for cortisol assays. We excluded 3 mother-infant dyads missing key covariate information, resulting in a final analytic sample of N=226 to examine joint effects of in utero exposure to metals and cortisol on infant temperament. Written informed consent was obtained from women prior to study participation in their preferred language. Study protocols were approved by the Institutional Review Boards (IRBs) of the Brigham and Women’s Hospital and the Icahn School of Medicine at Mount Sinai. Compared to the analytic study sample, mothers enrolled in PRISM who were excluded due to unavailable metal exposure and/or hair sample for cortisol, had a smaller proportion that identified as White, Non-Hispanic (13.9% vs 29.2%, p-value=0.004) and a greater proportion had less than high school diploma (17.3% vs 24.3%, p-value=0.02) (Table e1, online supplement).

Trace Elements

Women collected a spot urine sample in their home on the morning of a scheduled clinic visit in mid to late pregnancy. Samples were kept frozen in the participant’s freezer until transport to the PRISM laboratory. Immediately upon arrival, the urine samples were thawed, aliquoted, and stored at −80°C. Samples (200 μL) were diluted to 10 ml with a solution containing 0.05% Triton X-100, 0.5% nitric acid, and mixed internal standard before analysis of metal concentrations on an inductively coupled plasma-mass spectrometer-triple quadrupole (ICP-MS) instrument (Agilent 8800-QQQ). In this analysis we focused on seven trace elements that are non-essential for development: As, Ba, Cd, Cr, Cs, Pb, and Sb. Urine creatinine was measured using a well-established colorimetric method (limit of detection [LOD]: 0.3125 mg/dL)(Taussky, 1954). Quality control measures for trace elements and creatinine analysis have been previously described (Cowell et al., 2020). We replaced trace element concentrations below the LOD with the LOD divided by the square root of two. Trace element measures were log 10 transformed to address skewness. These approaches are likely to have minimal impact on our models given that exposure levels are converted to quantiles in our subsequent mixture analysis.

Hair Cortisol

Collection and analyses of hair cortisol in PRISM has been previously described (Bosquet Enlow et al., 2019). Briefly, research staff collected a 3 mm in diameter sample from the posterior vertex of the scalp (around 50 strands) within 1 week of delivery and samples were stored in opaque envelopes at room temperature out of direct sunlight. Each one-centimeter length of hair approximates one month of cortisol accumulation, with the hair closest to the scalp representing the most recent time-period (D’Anna-Hernandez et al., 2011). Hence, each sample was partitioned into 3-cm segments to represent pregnancy trimesters, with the third trimester being the segment closest to the scalp. Cortisol output during the third trimester was used in these analyses given prior research demonstrating that cortisol in late pregnancy may have the greatest influence on infant negative affectivity (Zijlmans et al., 2015). Additionally, focus on 3^rd trimester cortisol levels minimized exclusion of participants whose hair was too short to allow sampling for earlier trimesters. We have also previously demonstrated that correlations of hair cortisol across pregnancy trimesters were high (0.93 for adjacent trimesters and 0.86 for levels between the 1st and 3rd trimesters) in this sample (Schreier et al., 2016), Samples were measured by the Kirschbaum lab (Dresden, Germany) using enzyme-linked immunosorbent assay (ELISA) (n=92) or liquid chromatograph-spectrometry (LC-MS) (n=134) (Gao et al., 2013; Schreier et al., 2016). ELISA samples were converted into standard LC-MS equivalents using an established validated approach converting concentrations determined by each measure into assay-specific z-scores to account for residual variation introduced by assay type as previously described (Russell et al., 2015). Hair cortisol levels were transformed using a natural log transformation and dichotomized at the median (low vs. high).

Negative Affectivity

When infants were approximately 6 months of age, mothers completed the 191-item Infant Behavior Questionnaire—Revised (IBQ-R) designed to measure temperament between the ages of 3 and 12 months (Gartstein & Rothbart, 2003). The IBQ-R was designed to minimize reporter bias by inquiring about concrete infant behaviors rather than asking for abstract judgments. Mothers were asked to rate the frequency with which their infant demonstrated listed behaviors during the previous week on a 7- point scale from 1 (never) to 7 (always). The current study focused on the IBQ-R global factor of Negative Affectivity, which comprises four sub-domains – Fear, Sadness, Distress to Limitations, and Falling Reactivity/Rate of Recovery from Distress (hereafter Falling Reactivity). As previously described (Campbell et al., 2020; Cowell et al., 2021), the 16-item Fear scale assesses inhibited approach to novel objects or other stimuli, as well as distress in social situations that involve novelty or uncertainty; higher scores indicate greater fearfulness. The 14-item Sadness scale assesses general low mood and activity related to personal suffering, object loss, or inability to perform a desired action; higher scores indicate greater propensity to Sadness. The 16-item Distress to Limitations scale assesses the propensity to distress during caretaking activities or when confined by place or position; higher scores indicate greater tendency toward distress. Finally, the 14-item Falling Reactivity subscale assesses the rate of infant recovery from peak arousal, distress, or excitement and reflects the infant’s own abilities to regulate state; higher scores indicate more rapid recovery from arousal. The global composite of Negative Affectivity was calculated by taking the mean value of the four sub-scale scores, with Falling Reactivity being reverse-coded; thus, higher scores indicate greater Negative Affectivity. The IBQ-R has demonstrated good internal consistency, reliability, and validity in samples from different cultures as well as lower income samples (Parade & Leerkes, 2008), and our group confirmed the factor structure in our multi-ethnic sample (Bosquet Enlow et al., 2016).

Statistical Analyses

All statistical analyses were performed using R version 3.6.2. Descriptive statistics were calculated for covariate, exposures, and outcome measures (see Table 1). Internal consistency statistics were calculated for each subscale and the composite score, all indicating good internal consistency (Cronbach’s alpha ≥0.79, see Table e3). We used weighted quantile sum (WQS) regression to estimate the joint association of the mixture of seven trace elements and hair cortisol levels with each measure of infant Negative Affectivity (Negative Affectivity [global composite score], Fear, Sadness, Distress to Limitation, Falling Reactivity) in separate models. WQS is a mixtures-based ensemble modeling strategy that tests for associations between the combined effect of multiple exposures and an outcome of interest (Carrico et al., 2015). The WQS calculates a mixture index comprising the weighted sum scores of quantiled mixture components to test the association between the mixture of correlated metals and the outcome variables. Herein we used quintiles as it better captured the variability of exposure in our sample as done in prior work (Levin-Schwartz et al., 2022; Merced-Nieves et al., 2022). The WQS estimates a mixture effect in one direction at a time, which is done to avoid canceling out of mixture associations in the case of competing, bidirectional mixture effects. Given our focus on metals previously linked to poorer neurodevelopmental outcomes with higher exposure (a positive association), we constrained the WQS in the positive direction. Higher scores on the WQS index reflect higher exposure to metals. In the training dataset, weights were simultaneously estimated across bootstrapped samples (b=50) based on the association of each component (As, Ba, Cd, Cr, Cs, Pb, and Sb) with the outcome. Then this weighted index was tested in a linear regression model predicting the association between the mixture and the outcome. In order to address generalizability of the estimates across training and testing sets (80% and 20%, respectively), we implemented repeated holdout validation to randomly partition the data 100 times and repeat WQS regression on each set to produce a distribution of validated results (Tanner et al., 2019). The method additionally provides weights for each metal component, which convey the extent to which each contributes to the association between the mixture and the outcome. To evaluate the extent to which the metal components driving the association with infant Negative Affectivity and related features were dependent on maternal prenatal cortisol production and infant sex, product terms were introduced into the WQS model for each measure.

Table 1.

Sample characteristics for PRISM analysis subsample and by infant sex

	Analysis subsample (n = 226)		Boys (n = 123)		Girls (n = 103)
	N (%)	Mean (SD)	N (%)	Mean (SD)	N (%)	Mean (SD)
Maternal age		30.7 (5.6)		30.8 (5.5)		30.6 (5.9)
Race and ethnicity 1
White, non-Hispanic	66 (29.2)		36 (29.3)		30 (29.1)
Black	57 (25.2)		34 (27.6)		23 (22.3)
Hispanic	89 (39.4)		44 (35.8)		45 (43.7)
Other2	14 (6.2)		9 (7.3)		5 (4.9)
Maternal education
< High School	55 (24.3)		27 (22)		28 (27.2)
≥ High School	171 (75.7)		96 (78)		75 (72.8)
Gestational week at urine		32.3 (4.8)		32.4 (4.7)		32.2 (4.9)
Urinary metal (ng/mL)
Arsenic		23.9 (47.0)		23.7 (43.7)		23.7 (50.9)
Barium		4.2 (3.6)		4.5 (3.8)		3.9 (3.2)
Cadmium		0.2 (0.2)		0.2 (0.2)		0.2 (0.2)
Chromium		1.8 (2.8)		1.8 (2.0)		1.9 (3.6)
Lead		0.7 (0.6)		0.7 (0.5)		0.7 (0.7)
Antimony		0.1 (0.1)		0.1 (0.1)		0.1 (0.1)
Cesium		5.2 (3.4)		5.1 (3.3)		5.2 (3.5)
3rd trimester Hair cortisol
Low	113 (50.0)			55 (44.7)		58 (56.3)
High	113 (50.0)			68 (55.3)		45 (43.7)

¹Statistical models contained a dichotomized version of maternal race (White, non-Hispanic vs. Black/Hispanic/other).

²Participants classified as “Other” included those that self-identified as Asian (n=9), Multiple Race (n=4), and Other (n=1).

Covariates

For multivariable-adjusted models, we considered covariates associated with the exposures (metals, cortisol) and negative affectivity, but not on the causal pathway and formulated a directed acyclic graph (see figure e1) (Hernán et al., 2002) to determine the minimal sufficient adjustment set. These covariates were gestational week of urine collection, urinary creatinine, maternal race and ethnicity, prenatal maternal depressive symptoms, maternal education and age. Maternal race and ethnicity and education were self-reported at baseline; for the purpose of these analyses, these variables were dichotomized [non-Hispanic White vs. Black/Black-Hispanic, Hispanic/Non-black, and other race (hereafter Black/Hispanic//other race) and < high school degree vs. ≥ high school degree, respectively]. Prenatal maternal depressive symptoms were assessed by the validated Edinburgh Depression Scale (Murray & Cox, 1990). Infant’s age was not included in the models, as all infants were approximately 6 months of age (M=6.4, SD=0.3, Min=5.8, Max=7.3), hence the variability related to typical changes in temperament that occur throughout the first year of life was not of concern (Carnicero et al., 2000). We also adjusted for gestational week of urine collection (continuous in weeks) and urinary creatinine (continuous, mg/dL) to account for variability arising from pregnancy-related physiological changes and urine dilution that may influence metals concentrations.

Results

Descriptive data

The distribution of covariates in the analytic sample overall and by infant sex are summarized in Table 1. In the analytic study sample, the women were on average 30 years old at delivery, and the majority self-identified as Black or Black-Hispanic, Hispanic/non-Black, or other race (70.8%); 75.7% of mothers had at least a high school degree. Maternal urine was collected at 32.3 ± 4.8 weeks gestation. Urinary metal concentrations were positively correlated with Spearman coefficients ranging from 0.15 to 0.59 (P ≤ 0.0001–0.03) (see Table e2 in online supplement). Urinary concentrations of As (23.9 ± 47.0 ng/mL), Ba (4.2 ± 3.6 ng/mL), Cd (0.2 ± 0.2 ng/mL), Cr (1.8 ± 2.8 ng/mL), Pb (0.7 ± 0.6 ng/mL), Sb (0.1 ± 0.1 ng/mL), and Cs (5.2 ± 3.4 ng/mL). Third trimester hair cortisol had median levels of 0.84 pg/mg.

Metal Mixture

As illustrated in Figure 1, multivariable WQS regression models identified significant associations between the metals mixture and two of the outcomes –Sadness and Falling Reactivity. In the Sadness model, there was a 0.08 (95% CI = 0.06, 0.10) point increase with each quintile increase in the metals mixture. In the Falling Reactivity model, there was a 0.05 (95% CI = 0.03, 0.07) point increase with each quintile increase in the metals mixture. We did not observe a significant effect of the metal mixture in the Negative Affectivity (β = −0.007; 95% CI = −0.20, 0.19), Fear (β = −0.001; 95% CI = −0.34, 0.34). In the Distress to Limitations model (β = −0.05; 95% CI = −0.03, −0.07), the estimates were in the opposite direction of our hypothesis and the way the WQS was modeled. Also, when we modeled the WQS in the negative direction, there were no significant associations with any outcomes suggesting that the findings for Distress to Limitations was unstable and likely spurious.

Figure 1.

Changes in Negative Affectivity, Distress to Limitation, Fear, Sadness, and Falling Reactivity for each quintile increase in the metal mixture exposure Weighted Quantile Sum Index

Figure 2e displays the weighted quantile sum mixture weights for the significant metals mixture models. For the Sadness outcome, Ba the highest weights in the mixture, contributing 57.4%. For the Falling Reactivity model, Ba and Cd had the highest weights in the mixture contributing 41.5%and 23.6%, respectively.

Metal Mixture, Cortisol, and Infant Sex: Interactions

No two-way interactions between the metal mixture index and cortisol were significant (see figure e3). Figure 2 summarizes findings from models examining associations among the prenatal metal mixture, third trimester maternal hair cortisol levels, and infant sex in relation to the composite Negative Affectivity score as well as the separate subscale scores. Findings for boys and girls are shown in separate panels for illustrative purposes. As shown in Figure 2, multivariable WQS regression models identified significant interactions among the metal mixture, hair cortisol, and infant sex for two of five outcomes – Fear and Sadness. In the Fear model, girls born to mothers with higher cortisol had a 0.20 (95% CI=0.08, 0.32) score increase with each quintile increase of the metal mixture. In the Sadness model, girls born to mothers with higher cortisol had a 0.17 (95% CI=0.07, 0.23) score increase with each quintile increase of the metal mixture. We did not observe a significant 3-way interactions in the Falling Reactivity model (β = −0.07; 95% CI = −0.18, 0.04).

Figure 2.

Changes in Negative Affectivity, Distress to Limitation, Fear, Sadness, and Falling Reactivity for each quintile increase in the metal mixture exposure Weighted Quantile Sum Index by hair cortisol level and infant sex.

Figure 3 displays the weighted quantile sum mixture weights for the significant 3-way interaction models. For the Fear outcome, As, Cd, and Cs had the highest weights in the mixture, contributing 35.3%, 19.5%, and 20.0% respectively. For the Sadness outcome, Ba and Cr had the highest weights in the mixture, contributing 54.5% and 24.9%, respectively.

Figure 3.

Weight uncertainty plots depicting weighted quantile sum mixture weights (y-axis) for each of 100 repeated holdout validation sets for the temperament features that detected with significant 3-way interaction models. Abbreviations: As = Arsenic, Ba = Barium, Cd = Cadmium, Cr = Chromium, Cs =Cesium, Pb = Lead, Sb = Antimony. Notes: The magenta lines indicate the concern threshold of 25% in 100 repeated holdouts. Data points indicate weights for each of the 100 holdouts. Box plots show 25^th, 50^th, and 75^th percentiles, and whiskers show 10^th and 90^th percentiles of weights for the 100 holdouts. Closed diamonds show mean weights for the 100 holdouts.

Discussion

Summary of Results

Joint exposure to a mixture comprised of seven metals and elevated maternal cortisol in pregnancy were associated with infant Negative Affectivity attributes in a sexually dimorphic manner. Specifically, associations were stronger in relation to the Fear and Sadness temperament features in 6-month-old infants. The interaction in these models showed an increase in scores with higher exposure to the metal mixture in infants born to mothers with higher cortisol, particularly among girls. These findings corroborate numerous animal studies showing that exposure to metals and prenatal stress, or biological correlates of stress (i.e., hypercortisolism), can synergistically enhance CNS toxicity, and that effects are frequently sex dependent (Cory-Slechta et al., 2010; Cory-Slechta et al., 2012). These findings suggest that studies of the sexually-dimorphic neurotoxic effects of metals and psychological stress that consider one exposure in isolation will not reveal the full scope of their adverse effects on early neurobehavioral outcomes nor identify those at greatest risk for psychological problems in later life. Moreover, while it is not straightforward to draw comparisons between our findings in this human study considering impacts on infant temperament with prior animal research considering other neurobehavioral domains, these findings further highlight the need to consider sexually dimorphic effects.

Links between temperament features and later socioemotional difficulties have been reported in the child development literature. The domains of Fear and Sadness in infancy have been linked with later psychopathology, even though the magnitude of these associations have been variable (Abulizi et al., 2017; Kostyrka-Allchorne et al., 2020; Sayal et al., 2014; Slagt et al., 2016; Tang et al., 2020; Wu et al., 2022). Children higher in fearfulness appear particularly at risk for developing anxiety, whereas those higher in Sadness can be at heightened risk for developing depression (De Pauw & Mervielde, 2010). The IBQ-R Fear scale captures behavioral inhibition; a temperament trait characterized by startle or distress to sudden changes in the environment and inhibited approach to novelty. Our finding of an increase in Fear scores with higher exposure to the metal mixture in infants born to mothers with higher cortisol suggests a multiplicative effect of the metal mixture and cortisol on behavioral inhibition, which has been shown to be an early precursor of anxiety disorders (Cowell et al., 2021). Notably, this finding was most evident among girls. In general, females have been shown to be more vulnerable to the psychological consequences of stress, such as depression and anxiety, in that they are generally seem to be at higher risk for developing these disorders (Hankin et al., 2007; Seedat et al., 2009; Tolin & Foa, 2006). It is important to note that other factors such as culture and race and ethnicity also contribute to sex-specific vulnerability to psychopathology (Hong et al., 2022).

Potential Mechanisms

Several lines of overlapping evidence suggest mechanisms that may underlie the observed interactions. Prenatal exposure to metals and cortisol are hypothesized to impact behavioral development through their shared biological targets, specifically the developing HPA axis (Cory-Slechta et al., 2017) and the brain mesocorticolimbic dopaminergic system, which includes the prefrontal cortex (PFC) and hippocampus , highly interactive brain regions involved in the programming of neurobehavioral outcomes (Burke & Miczek, 2014). Brain development occurs sequentially over gestation, with different anatomic regions forming in a timed cascade (Sunyer & Dadvand, 2019). Several structural components and related processes can be differentially susceptible to environmental toxicants linked to differing domains of neurodevelopment. Starting in utero, toxic metals and psychological stress, and their physiological correlates such as hypercortisolism, disrupt similar but not completely overlapping neural processes that may underlie an unfolding predisposition to internalizing problems manifested across childhood – e.g., disrupted synaptic transmission (Dickerson et al., 2020), dopaminergic systems (Jones & Miller, 2008), and/or stress response systems, including the maternal-fetal HPA axis (Sobolewski et al., 2018). Our group and others have begun to show that brain regions implicated in emotion regulation and related cognitive control areas are vulnerable to metals (de Water et al., 2019; de Water et al., 2018). The PFC, anterior cingulate cortex (ACC), and insula seem particularly vulnerable to prenatal metal exposures (Gump et al., 2017; Vänskä et al., 2019). Metal exposures have also been linked to negative emotionality and problems in early orientation and regulation (Fruh et al., 2019; Rocha & Trujillo, 2019).

Metals and stress exposures influence interrelated but not completely overlapping processes that may jointly impact sexually dimorphic brain development (Cowell & Wright, 2017). Differing structure volumes, neuronal morphology, and synaptic connections lead to sexually dimorphic brain circuitry (Cosgrove et al., 2007; Schwarz & McCarthy, 2008). Starting in utero, sex hormones have organizational effects on the development of brain structures that control sexually dimorphic neuroendocrine responses and behaviors (Krolick & Shi, 2022). Interactions between the HPA axis and the hypothalamic-pituitary-gonadal (HPG) axis may underlie sex-specific effects on behavior. Chemical and non-chemical stressors, including metals, disrupt HPA-HPG signaling during early development (Izvolskaia et al., 2016; Xie et al., 2020), impacting brain development (Cowell & Wright, 2017). The HPG axis has been shown to be affected by chronic activation of the HPA axis, which may result from chronic stress. The production of cortisol inhibits the release of gonadotropin releasing hormone, which, in turn, affects the rest of the pathway and the production of sex steroids (estrogen in ovaries and testosterone in testes) (Toufexis et al., 2014). Surges of gonadal and adrenal testosterone are important for the sexual differentiation of the brain (Cowell & Wright, 2017). Finally, upregulation of the HPA axis may lead to greater production of androgens by the adrenal cortex, and this may influence the normal sex differentiation of the brain (Barrett & Swan, 2015; Cowell & Wright, 2017). Glucocorticoids and sex steroids differentially influence kinetics and toxicity of metals in males and females (Schwarz & McCarthy, 2008). Sex-specific properties of the blood brain barrier result in differences in toxicant uptake and elimination (Zhao et al., 2022). Sex differences in antioxidant defense, metabolizing enzymes, and placental responses also play a role (Bale, 2016; Lesseur et al., 2014).

Prenatal exposure to the metal mixture and increased cortisol levels were jointly associated with infant Fear and Sadness temperament features, with effects stronger among girls. The temperament characteristics of Fear and Sadness are hypothesized to be dependent upon co-occurring maturation of critical social, emotional, and regulatory brain regions (Kaczkurkin et al., 2019), sharing localization in affective networks encompassing the amygdala, insula, hippocampus, ACC, and PFC (Robinson et al., 2014). Prior research has implicated a role for amygdala-ventromedial prefrontal cortex (vMPFC) connectivity in anxiety and depression (Casey & Lee, 2015; Connolly et al., 2017), with emerging research demonstrating individual differences in amygdala connectivity relevant to the expression of Fear and Sadness evident at birth, with connections relevant for Fear development distinct from those predicting Sadness trajectories (Thomas et al., 2019). Also, although amygdala-vMPFC connectivity has been linked with pathological and subclinical variation in Fear (Baur et al., 2013), it has more consistently been associated with depressive psychopathologies (Perlman et al., 2012). Alterations in functioning of these interconnected regions that occur as a result of interactions between metals and elevated cortisol exposure in utero may predispose individuals to the development of affective disorders.

Strengths and Limitations

This study has several strengths. First, to our knowledge, this the first human study showing that the joint association of prenatal exposure to a metal mixture and higher maternal cortisol levels in pregnancy is related to infant Negative Affectivity in a sex-specific manner. Consideration of metal mixtures addresses the practical reality that exposures experienced in ‘real life’ involve multiple chemicals simultaneously (Merced-Nieves et al., 2021). Specifically, we implemented WQS regression analysis, a supervised machine-learning technique, to examine comprehensively the mixed effects of multiple, correlated metals on infant Negative Affectivity. The WQS method groups each exposure variable into quantiles, which protects against potential misestimating of effects due to outliers or extreme values, and provides an across-scale normalization. The racially/ethnically and socioeconomically diverse sample also is a strength given that racially and ethnically underrepresented populations and lower SES communities are more burdened by the toxic exposures being considered.

There are also some limitations that should be considered. Generalization may be limited to populations experiencing similar developmental and environmental contexts to this cohort. Metal concentrations were measured in maternal urine, which is not the preferred matrix for all considered metals. Urine is considered a more optimal matrix for measuring As, Ba, Cd, Cs, Cr, and Sb, while whole blood is considered the gold-standard matrix for Pb. Moreover, while urinary Cd and Cs are considered acceptable biomarkers of long-term exposure, urinary concentrations of the other metals reflects recent exposure and thus may not provide a measure of average exposure across the course of pregnancy when using a single measurement. Given the variable toxicokinetics of the different metals contributing to the mixture, no single biosample medium can fully capture profiles of metals in a mixture. Thus, future studies that consider metals measured in blood or other matrices as well as studies with repeated biosample measures would provide a more comprehensive picture of prenatal exposure. Future research that integrates exposure biomarkers across multiple matrices (e.g., Cd in urine, Pb in blood) will advance this area of inquiry and is needed to further substantiate our findings. Additionally, there could be exposure misclassification in our measure of As, as total urinary As reflects both toxic and nontoxic species. With regard to our modeling approach, WQS regression includes a directionality constraint, such that the examined mixtures consist of metals either all positively or negatively associated with the outcome. Given this constraint, we restricted our models to include only those metals we hypothesized to be associated with worse negative affectivity based on prior neurodevelopmental research. However, we note that many metals are essential micronutrients that play important roles in neurodevelopment. Future work should consider the combined effect of toxic and essential elements, which may act in opposing directions.

Temperament outcomes were based on maternal report. The IBQ-R is a validated and widely used approach to assessing infant temperament in epidemiologic studies and offers the strength of providing information from caregivers who have the opportunity to observe children over long periods of time and in multiple contexts. The measure was designed to reduce the influence of reporter biases by inquiring about concrete child behaviors rather than asking for abstract judgments and has demonstrated strong psychometric properties (Gartstein & Rothbart, 2003). In addition, the IBQ-R presents sets of items based on the context or situation eliciting the infant’s reactions (e.g., bathing, dressing), which serves to enhance specific recall and limit social desirability. Future work may consider incorporating independent observational assessments of child temperament to more fully explicate associations among metal mixtures, prenatal cortisol exposure, and child temperament. While we adjust for prenatal depressive symptoms, we do not consider concurrent depressive symptoms to maternal reporting of child temperament. Although recent research suggests that maternal psychopathology likely has small to no impact on biasing mothers’ reporting on their child’s behavioral problems (Olino et al., 2020; Olino et al., 2021) , we cannot rule out this possibility altogether. We focus on the mother as the reporting primary caregiver and future work should also consider fathers. We focus on the mother as the reporting primary caregiver and future work should also consider fathers. Finally, another limitation implicit to the analysis of early-life exposures is the potential impact of uncontrolled variables, including unmeasured exposures (e.g., seasonality, medication use). Also, prenatal exposure to toxins such as metals and elevated cortisol can be modified by postnatal factors that build resilience in children. Future studies should consider how the postnatal caregiving environment may modify the effects of such prenatal exposures (Nolvi et al., 2023). Future studies with larger samples in socioeconomically and/or racially and ethnically diverse samples would also allow more detailed assessment on how these associations may vary across racial/ethnic groups or in relation to socioeconomic status.

Conclusions

The current findings underscore the need to consider complex interactions among prenatal exposure to multiple metals, biomarkers of disrupted stress reactivity in utero, and infant sex in order to more fully elucidate the intergenerational effects of metal exposures on children’s neuropsychological development. This study also highlights the critical need to broaden screening programs for exposure to health-relevant metals, currently restricted to young children, to pregnant at-risk women.